1
DOES PAROLE WORK?
t
Analyzing the Impact of Postprison Supervision on Rearrest Outcomes
RESEARCH HIGHLIGHTS

Overall, parole supervision has
little effect on rearrest rates of
released prisoners. Mandatory
parolees, who account for the
largest share of released prison-
ers, fare no better on supervi-
sion than similar prisoners
released without supervision. In
fact, in some cases they fare
worse. While discretionary
parolees are less likely to be
rearrested, this difference nar-
rows (to 4 percentage points)
after taking into account per-
sonal characteristics and crimi-
nal histories.
(Continued on page 2)
MARCH 2005
AMY L. SOLOMON
VERA KACHNOWSKI
AVINASH BHATI
URBAN
INSTITUTE
2100 M STREET, N.W.
WASHINGTON, D.C. 20037
he vast majority of prisoners in this country (about 80 percent) are
released “conditionally,” subject to a period of supervision in the
community, often called “parole.”1 Parole supervision is used as both
a surveillance tool and a social service mechanism and ideally serves a
deterrent role in preventing new crimes from occurring. Parole
supervision can function as a surveillance tool by monitoring and
sanctioning those who violate conditions of release, potentially averting
more serious reoffending. Parole supervision can also act as a social service
mechanism by using rules and incentives to engage ex-prisoners in positive
activities, such as work and drug treatment, and to place ex-prisoners in pro-
grams that may help reentry transitions. While the focus of parole supervision
has shifted more toward the surveillance function over the years,2 the number
of people subject to it continues to grow. In 2003, over 774,000 adult men and
women were under parole supervision in the United States,3 up from 197,000
in 1980.4
Despite its widespread use, remarkably little is known about whether parole
supervision increases public safety or improves reentry transitions. Prior
research indicates that fewer than half of parolees successfully complete their
period of parole supervision without violating a condition of release or commit-
ting a new offense,5 and that two-thirds of all prisoners are rearrested within
three years of release.6 To date, however, no national studies have compared the
criminal activity of prisoners who are supervised after release to that of their
unsupervised counterparts.
In this research brief, we use data from a Bureau of Justice Statistics (BJS) recidi-
vism study7 (see “Data Sources and Limitations” sidebar) to compare prisoners
released to parole supervision in 1994 with prisoners who completed their
entire prison sentence and were released without any supervision or reporting
requirements.8 Our goal is to assess, at an aggregate level, whether parole
“works” at reducing recidivism among those who are supervised after release
from state prison.
2
RESEARCH HIGHLIGHTS
(Continued from page 1)

Certain prisoners benefit more from supervision—espe-
cially discretionary release to supervision—than others.
For example, females, individuals with few prior arrests,
public order offenders, and technical violators are less
likely to be rearrested if supervised after prison. Persons
with a combination of these characteristics, representing
relatively low-level offenders, exhibit even lower rearrest
rates if supervised. Conversely, supervision does not
The report is organized around three key questions.
First, do prisoners released with and without supervision
differ with respect to demographics, incarceration char-
acteristics, and criminal histories? Second, do prisoners
released with and without supervision recidivate at dif-
ferent rates? And finally, if there are differences in recid-
ivism outcomes between those released with and without
supervision, when and for whom does supervision mat-
ter most?
BEYOND CONDITIONAL AND
UNCONDITIONAL: THE ROLE OF
THE RELEASE MECHANISM
To assess the relationship between parole supervision
and recidivism, we must look not only at whether a
person is supervised after release, but also how they
were released. Persons released unconditionally—with-
out any postrelease supervision—are released when
their sentences end. Alternatively, persons released con-
ditionally are released to supervision by two different
methods, discretionary release and mandatory release.
In this section we describe the differences between these
release mechanisms and examine shifting trends in
their use.
Prisoners released to supervision via discretionary release
have been screened by a parole board or other authority
improve rearrest outcomes for some of the higher rate,
more serious offenders.

Of the largest groups of released prisoners—male drug,
property, and violent offenders—only property offenders
released to discretionary parole benefit from supervi-
sion. Violent offenders released to supervision are no
less likely to be rearrested than their unsupervised coun-
terparts. For male drug offenders, mandatory release to
supervision predicts higher rearrest rates than for uncon-
ditional releasees or discretionary parolees.
to determine “readiness” to return to the community.
Parole boards, which often face substantial pressures to
reduce prison overcrowding, determine who presents
the lowest risk of reoffending and is most prepared for
release. Among other factors, parole boards consider
criminal histories, institutional conduct, and positive
connections to the community such as employment,
housing arrangements, and ties to family. Appearing
before a parole board may provide an incentive for pris-
oners to participate in programming and arrange tran-
sition plans to improve their chances of early release.
Until the 1980s, discretionary parole was the predomi-
nant method of release, accounting for 55 percent of all
prison releases. Over the past two decades, however,
discretionary release has largely fallen out of favor with
policymakers.9 By 2000, just 24 percent of released pris-
oners were discretionary releasees (figure 1),10 and 16
states had abolished discretionary release altogether.11
Other states have retained discretionary release but limit
its use to certain offenses.12
Mandatory release to supervision typically occurs in
states that use determinant sentencing schemes and now
accounts for about 40 percent of all prison releases.13
Mandatory release occurs when a prisoner has served
his original sentence, less any accumulated good time
credit, and serves the remaining balance of his sentence
under supervision in the community. Good time credit

DATA SOURCES AND LIMITATIONS
This report relies primarily on Bureau of Justice Statistics
(BJS) data on 38,624 prisoners released in 1994 from
15 states: Arizona, California, Delaware, Florida, Illinois,
Maryland, Michigan, Minnesota, New Jersey, New York, North
Carolina, Ohio, Oregon, Texas, and Virginia. This sample of
prisoners is representative of the 272,111 prisoners re-
leased from those states in 1994—two-thirds of all prisoners
released nationwide in 1994. Due to issues with the data,
Delaware is excluded from the analysis in this report. The BJS
data used for this analysis are available at the Inter-University
Consortium for Political and Social Research (ICPSR) website,
http://www.icpsr.umich.edu. The full BJS report, “Recidivism
of Prisoners Released in 1994,” by Patrick Langan and David
Levin, is available at http://www.ojp.usdoj.gov/bjs/abstract/
rpr94.htm. Our analysis also used Census Bureau data avail-
able at http://www.census.gov/main/www/cen1990.html
and National Corrections Reporting Program (NCRP) data,
available at the ICPSR website.
There are several limitations to our data. First, our analysis
relies on rearrests as a proxy for reoffending, although
rearrests reflect a combination of both criminal activity and
policy decisions (e.g., to report a crime, to arrest an individ-
ual, to revoke parole). Further, those on supervision may
be watched more closely by law enforcement as well as
parole officers, and thus criminal activity committed by
is typically earned through program participation or
good behavior while incarcerated. Mandatory releasees
have not received a determination of fitness to return to
the community from a parole board or other author-
ity.14 Postprison supervision resulting from discre-
tionary or mandatory release is not systematically
different. In most states, conditions of supervision are
similar for both types of parolees, although discre-
tionary parolees often spend more time under supervision
than mandatory parolees (see “State Variation” sidebar).15
Finally, unconditional release occurs when prisoners
have served the entirety of their sentence behind bars
and must be released without any conditions, commu-
nity supervision, or reporting requirements. As such,
these individuals cannot be returned to prison for any
3
parolees may be more likely to be detected than criminal
activity commited by unconditional releasees. Future
research would benefit from self-reported data on actual
offending behavior.
Additionally, while the BJS data sources provided important
information on the personal and criminal histories of re-
leased prisoners, information on the nature of supervision
was not available. Individual-level data on intensity of super-
vision, length of supervision, reporting requirements, and ser-
vices received would be useful. And although we were able to
control for state-level effects, we did not have specific infor-
mation on differences in state sentencing and parole prac-
tices and revocation policies. In particular, the research
would benefit enormously from system-level data about risk
assessment tools, contact standards, caseload averages,
case planning, case management strategies, and neighbor-
hood-based supervision models. Without such information,
we are unable to consider how various types of supervision
affect rearrest outcomes. For example, perhaps some super-
vision strategies are very effective but the aggregate level of
the data does not allow observation of these differences.
Finally, as discussed in the “State Variation” sidebar, parole
practices and outcomes vary substantially across states.
The aggregate nature of our analysis may bury significant
differences at the state level relative to the outcomes
associated with parole.
FIGURE 1. Share of State Prisoners Nationwide Released
Conditionally and Unconditionally, 1980–2000
Percent
60
50
Conditional, discretionary
40
30
Conditional, mandatory
20
Unconditional
10
0
1980 1984 1988 1992 1996 2000
Source: Hughes, Wilson, and Beck, “Trends in State Parole,
1990–2000” and National Corrections Reporting Program.
4
DEFINITIONS OF KEY TERMS
Conditional Release: Release from prison to supervi-
sion with a set of conditions for remaining on parole,
which, if violated, can cause the person to be
returned to prison. This subsequent incarceration
can be for any of the remaining portion of the sen-
tence the inmate may have on the current offense.
Determinate Sentencing: Fixed prison terms that
can only be reduced by good time or other earned
time-reduction credit.
Discretionary Release to Parole Supervision:
Prisoners are conditionally released to supervision
based on a statutory or administrative determination
of eligibility by a parole board or other authority.
Good Time Credit: Credit earned by prisoners that
reduces their total length of stay in prison. Good time
credit may be awarded for good behavior, program
participation, exceptional deeds, or, in some cases,
automatically. Except for Hawaii, Montana, and Utah,
all states have a good time credit system.
Indeterminate Sentencing: A sentencing structure,
common in the early 1970s, where no fixed term is
assigned and parole boards are given the authority to
release offenders from prison.
Mandatory Release to Parole Supervision: Prisoners
are conditionally released to supervision after serving
a portion of their original sentence less any good time
credit earned. Mandatory release generally occurs in
jurisdictions using determinate sentencing statutes.
portion of their original sentence (e.g., their release
cannot be revoked for violating conditions of release,
because there are none). Similar to mandatory releasees,
prisoners released unconditionally were not granted
early release via a parole board in states retaining dis-
cretionary parole. Further, unconditional releasees did
not earn good time credit while incarcerated and were
imprisoned until their original sentence expired.
New Court Commitment: Persons entering prison
directly from a court sentence for a new offense, and
not from an unsuccessful period of community super-
vision (parole).
Parole or Probation Violator Commitment: Persons
entering prison as a result of a parole or probation
violation, such as violating a condition of supervision.
Parole Supervision: A period of conditional super-
vised release following a prison term. Prisoners may
be released to parole either by a parole board deci-
sion (discretionary) or according to statutory provi-
sions (mandatory).
Unconditional Release: Release from prison upon
the expiration of the sentence, without being subject
to any conditions of release or supervision in the
community. These persons have served their entire
prison term and thus do not face the possibility of
return to prison for the current offense.
Sources: Definitions adapted from Paula Ditton and Doris
Wilson. 1999. “Truth in Sentencing in State Prisons.”
Bureau of Justice Statistics Special Report. Washington,
DC: U.S. Department of Justice, Office of Justice Programs;
Timothy Hughes, Doris Wilson, and Allen Beck. 2001.
“Trends in State Parole, 1990–2000.” Bureau of Justice
Statistics Special Report. Washington, DC: U.S. Department
of Justice, Office of Justice Programs; Greg Jones, Michael
Connelly, and Kate Wagner. 2001. “The Effects of
Diminution Credits on Inmate Behavior and Recidivism: An
Overview.” College Park: Maryland State Commission on
Criminal Sentencing Policy.
Prisoners released without supervision account for
about one-fifth of all prison releases,16 with substantial
variation across states.
The vast majority of prisoners in the BJS recidivism
study were released conditionally: Mandatory releases
to supervision accounted for 57 percent of released
prisoners and discretionary releases accounted for
 5

35 percent. The remaining 8 percent of prisoners in the
sample were released unconditionally.17 Throughout
this report, we discuss released prisoners’ characteris-
tics and recidivism outcomes in terms of three cate-
gories of releasees: (1) prisoners released conditionally
following discretionary release, referred to as “discre-
tionary parolees”; (2) prisoners released conditionally
following mandatory release, referred to as “manda-
tory parolees”; and (3) prisoners released uncondition-
ally, referred to as “unconditional releasees.” While we
refer to prisoners released to parole as “parolees,” they
were not necessarily on parole throughout the entire
period of analysis.18
CHARACTERISTICS OF
RELEASED PRISONERS
We began our analysis with an examination of the re-
lease groups’ demographics and criminal histories, as well
as the characteristics of the incarceration from which the
prisoner was most recently released. Specifically, we were
interested in whether persons released conditionally (via
mandatory and discretionary release) and uncondition-
ally had different risk factors that would suggest a greater
likelihood of postrelease recidivism.19 Prisoners in all
three release categories had similar demographic charac-
teristics. The average age at release among all three cate-
gories was 32 or 33 years old, and the vast majority was
male. Just over half of unconditional releasees and discre-
tionary parolees were black, compared with about 40 per-
cent of mandatory parolees (table 1).
More than 90 percent of each group had been arrested
in the past. Unconditional releasees and mandatory
parolees, however, had slightly higher average numbers
of prior arrests than discretionary parolees. We also ex-
amined prior arrests for violent crimes as another indi-
cator of potential risk to the community upon release.
Our analysis showed that larger shares of prisoners
released unconditionally had previously been arrested for
a violent offense than had mandatory parolees, with dis-
cretionary parolees the least likely to have been arrested

TABLE 1. Characteristics of Prisoners Released in 1994, by Supervision Status at Release

Unconditional releasees Mandatory parolees Discretionary parolees

Demographics
Average age at release (years) 32.7 32.6 31.9
Male (%) 93 92 90
Black (%) 55 42 54

Criminal history
Previously arrested (%) 93 94 92
Average number of prior arrests 9.6 9.5 7.5
Previously arrested for violent offense (%) 67 63 55
Prior incarcerations (prison or jail, %) 68 69 67
Average number of prior incarcerations 2.7 2.5 2.3

Incarceration characteristics
Incarcerated for violent offense (%) 27 21 23
Incarcerated for drug offense (%) 30 31 34
Incarcerated for property offense (%) 33 35 31
Incarcerated for public order offense (%) 9 9 10
Average time served (months) 32.0 18.5 21.3

Source: Urban Institute analysis of Bureau of Justice Statistics data.

6
for a violent offense in the past (figure 2). Around two-
thirds of each group had been confined to prison or jail
previously, two to three times on average (table 1).
Additional Findings

A higher share of the prisoners released uncondition-
ally were serving sentences for a violent offense com-
pared with those released to supervision. A slightly
higher share of discretionary parolees were serving
time for a drug offense compared with the other
groups, while a slightly higher share of mandatory
parolees were property offenders (table 1).

Over two-thirds of discretionary parolees were serving
sentences for a new court commitment, compared
with about half of unconditional releasees and manda-
tory parolees. In other words, higher shares of the
unconditional releasees and mandatory parolees were
incarcerated most recently for a parole or probation
revocation.

Prisoners released unconditionally served almost a
year longer behind bars, on average, than prisoners
released to supervision. The longer terms served by
unconditional releasees may reflect the nature of their
charge or institutional conduct that prevented them
FIGURE 2. Share of Released Prisoners with Prior Arrests, by Supervision Status at Release
Percent
100
90 93 94 92
80
70
60
50
40
30
20
10
0
Any prior arrest
Source: Urban Institute analysis of Bureau of Justice Statistics data.
from earning good time credit.20 For those uncondi-
tional releasees who returned to prison for a parole
revocation, the longer time served may reflect their
serving the remaining balance of their original sen-
tence behind bars, which is typical in some states.
In sum, while individuals in all three release groups have
similar demographic characteristics, unconditional
releasees and mandatory parolees appear to be slightly
higher-risk populations overall, as compared with discre-
tionary parolees, given that they had more prior arrests
and were more likely to have previously been arrested for
a violent offense. Unconditional releasees were also
slightly more likely than their supervised counterparts to
be serving time for a violent offense. Further, the fact
that prisoners released unconditionally served longer
terms on average, and thus had been removed from soci-
ety for a longer period of time, suggests that they may
have become more disconnected from positive social
networks than their supervised counterparts.
RECIDIVISM OUTCOMES
The BJS recidivism study found that within two years
of release, 59 percent of ex-prisoners were rearrested,
Unconditional
Mandatory parolees
Discretionary parolees
67
63
55
Violent prior arrest

WHAT DOES PRIOR RESEARCH SAY ABOUT THE
IMPACT OF SUPERVISION ON RECIDIVISM?
While no national studies have compared rearrest rates for
U.S. prisoners released with and without supervision, a
small study investigated outcomes for prisoners released
in Canada in 1968 and found similar trends to those
described in this report.1 The study compared the rearrest
rates of discretionary parolees (n=210) to unconditional
releasees (those who had applied for parole and been
denied [n=100] and those who never applied for parole
[n=113]). Within two years of release, 68 percent of the
unconditional releasees were rearrested, compared with
44 percent of the parolees. When the study controlled for
the fact the parole boards select “better risk” individuals,
however, the differences in rearrest rates between parol-
ees and unconditional releasees were virtually identical.
The most rigorous study of supervision and recidivism in
the United States is a nine-state randomized evaluation
that compared offenders monitored in Intensive Super-
vision Programs (ISPs) to those subject to standard
supervision.2 The authors found little difference in overall
rearrest rates between the ISP treatment group and the
control group, although the treatment group had consider-
ably higher levels of technical violations—likely the result
of heightened surveillance inherent in ISPs. The study was
unable to determine whether intensive monitoring and
sanctioning of technical violations actually resulted in
improved public safety outcomes. The research did, how-
ever, show that intensive supervision was successful at
increasing program participation. Importantly, a review of
36 percent were reconvicted, and 19 percent were returned
to prison with new sentences.21 We expand on these recid-
ivism findings by describing rearrest outcomes by super-
vision status. We focus on rearrest as the closest proxy to
offender behavior, acknowledging that rearrest is an im-
perfect measure of the relationship between supervision
and criminal activity. Rearrests do not measure how much
actual reoffending has occurred, but how much criminal
activity has been detected, and supervision increases the
likelihood that criminal activity will be detected.22 While
7
four ISP studies, including the one described above, found
that supervision strategies that included some level of
rehabilitation or treatment in combination with surveillance
techniques were more effective in reducing rearrest rates
than surveillance alone.3
In addition, a small yet relevant study explored the impact
of probation on the criminal activity of 125 offenders.4 It
compared offenders’ outcomes while on probation with
their outcomes in the year preceding probation. The study
found that probation did have an impact on the criminal
activities of probationers, particularly among older offend-
ers and property and drug offenders. Probation appeared to
reduce the number of offenders who recidivated, the rate
of offending among recidivists, and high-risk behavior
linked to crime. It was not clear, however, whether the
arrest and/or sentencing events—and not probation super-
vision itself—actually affected offending behavior.
1
Irwin Waller. 1974. Men Released from Prison. Toronto, Ontario:
University of Toronto Press.
2
Joan Petersilia and Susan Turner. 1993. “Intensive Probation and Pa-
role.” In Crime and Justice: A Review of Research, vol.17, edited by
Michael Tonry. Chicago: University of Chicago Press. See also Joan
Petersilia and Susan Turner. 1993. “Evaluating Intensive Supervision
Probation/Parole: Results from a Nationwide Experiment.” NIJ Research
in Brief. Washington, DC: U.S. Department of Justice, Office of Justice
Programs.
3
Lawrence W. Sherman, Denise Gottfredson, Doris MacKenzie, John
Eck, Peter Reuter, and Shawn Bushway. 1998. Preventing Crime: What
Works, What Doesn’t, What’s Promising. Research in Brief. Washington,
DC: National Institute of Justice. http://www.ncjrs.org/works/.
4
Doris MacKenzie, Katharine Browning, Stacy Skroban, and Douglas
Smith. 1999. “The Impact of Probation on the Criminal Activities of Of-
fenders.” Journal of Research in Crime and Delinquency 36(4): 423–53.
the rearrest rates for all three groups may underestimate
the actual incidences of recidivism, it may be especially
true for unconditional releasees, as the absence of supervi-
sion reduces the likelihood that criminal activity will be
detected. That said, we still expected to find that prisoners
released without supervision would be rearrested more
frequently than conditional releasees given the characteris-
tics of the groups (discussed above), combined with the
absence of supervision (we assumed supervision would
deter some criminal behavior among parolees).
8

Our analysis indicates that in the two years after their
release, discretionary parolees were less likely to be
rearrested than both mandatory parolees and pris-
oners released unconditionally. Just over 60 percent
of unconditional releasees and mandatory parolees
were rearrested at least once over two years, compared
with 54 percent of discretionary parolees.23 Individuals
in each group had between two and two and a half
rearrests, on average, during the two-year period
(table 2).
We evaluated the safety threat these groups posed to the
community by examining the types of offenses for which
prisoners were rearrested. We started by looking at the
distribution of offenses among those who were rear-
rested at least once. Roughly the same shares of all three
groups were first rearrested for property offenses, while a
higher share of mandatory parolees were first rearrested
for drug offenses, and a slightly higher share of uncondi-
tional releasees were first rearrested for violent crimes
(figure 3). We then determined what percentage of all
prisoners in each group were arrested for a violent crime
in the two years following release. About one-fifth
(22 percent) of unconditional releasees were rearrested
for a violent crime during the two years following
release—a larger share than mandatory (17 percent)
or discretionary (14 percent) parolees.
TABLE 2. Rearrest Outcomes after Two Years, by Supervision
Status at 1994 Release
Unconditional Mandatory Discretionary
releasees parolees parolees
Percent rearrested 62% 61% 54%
Average number of
rearrests 2.5 2.1 2.1
Source: Urban Institute analysis of Bureau of Justice Statistics data.
Our finding that discretionary parolees were less likely
to be rearrested than unconditional releasees and
mandatory parolees could suggest that parole boards are
making sound decisions by choosing lower risk prison-
ers for release. Alternatively, it could indicate that
supervision has a more beneficial impact on this group
than on mandatory parolees.24 To isolate the impact of
supervision on rearrest, we conducted statistical model-
ing that controlled for all other demographic, criminal
history, and contextual variables included in the analysis
(see “Methodology” sidebar). The results revealed that
when comparing two individuals with similar char-
acteristics, their rearrest outcomes—based exclusively
on their supervision status—differ only slightly. Spe-
cifically, when all other variables were controlled for, the
predicted probability of rearrest for mandatory parolees
and unconditional releasees was identical at 61 percent,

FIGURE 3. First Rearrest Offense of Prisoners Who Were Rearrested at Least Once, by Supervision Status at 1994 Release
Percent
35
Unconditional releasees
30
31 30 Mandatory parolees
29
25 27
25 Discretionary parolees
20 23
21
19 19 19
15 17 16
10
10 10
5
5
0
Drug Property Violent Public order Other/Unknown

Source: Urban Institute analysis of Bureau of Justice Statistics data.


STUDY METHODOLOGY
The results presented in this report are based on a com-
bination of descriptive analysis, regression analysis, and
simulation results. We relied on descriptive analysis to
address the first two questions of interest—whom states
release conditionally versus unconditionally, and whether
these groups recidivate at differential rates. These results,
which are discussed in first half of the report, are based on
comparisons of the average characteristics among the
three categories of released prisoners. As such, the
results are used only to describe the three groups of
releasees. To address the remaining research question—
when and for whom does supervision matter most—we
relied on a combination of multivariate regression and
simulation analyses.
We used multivariate logistic regression analysis to better
understand the impacts that individual and community
characteristics may have on the probability of recidivism for
individuals released under supervision and those released
unconditionally. Unlike the necessarily bivariate nature of
the descriptive analysis, the multivariate analysis allows
us to control for the effects of all included characteristics
simultaneously. Our final models of the probability of an
individual’s rearrest within the two-year follow-up period
included (1) the individual-level characteristics shown in
table 3—criminal history, age, race, admission type, and
offense type; (2) an index of community indicators we
called “resource deprivation” that takes into account sev-
eral characteristics of the community to which the released
prisoner returns; (3) the prisoner’s supervision status at
time of release; and (4) a set of state-level fixed effects.
Some additional variables, which were initially included in
our models, were dropped from the analysis because they
were insignificant. Additionally, to allow the effects of
supervision status at time of release to vary across indi-
viduals and communities, we included a set of interaction
terms.
Our resource deprivation index (an empirically defined
index) was created from four county-level measures
obtained from the Census Bureau that were combined
using factor analysis. The four measures included in the
final score are (1) proportion of all households headed by
9
females; (2) county poverty rate; (3) county unemployment
rate; and (4) share of county population that is black. We
used county of sentencing as the best proxy for county of
return because the latter was unavailable.
Based on the results of the regression analysis, we esti-
mated probabilities of rearrest within a two-year follow-up
period for various combinations of the included characteris-
tics. This simulation analysis entails estimating and com-
paring predicted probabilities by varying one or more of the
included characteristics while holding the others fixed. In
this way, we are guaranteed to capture only the effects of
those characteristics that are varied (i.e., their marginal
effects). For discrete measures, such as race and gender,
this simulation analysis is natural and easy to perform. For
continuous variables, however, we need to select specific
points at which to simulate the probabilities. For the three
continuous variables included in our models—number of
prior arrests, age, and resource deprivation—we selected
three analysis points based on the distribution of these
variables in the data. These points represent typical “low,”
“medium,” and “high” values for each of these characteris-
tics. For prior arrests, three prior arrests represented a typ-
ical person with few prior arrests, six represented a typical
person with medium prior arrests, and 12 represented a
typical person with high prior arrests. For age, we used
25 to represent low, 31 to represent medium, and 37 to
represent high age. For resource deprivation, an index of
factor analysis scores was created, ranging from 0 to 100.
Based on the distribution of scores, a score of 31 repre-
sented low, 42 represented medium, and 47 represented
high. Using different points of evaluation for the simulation
changes the predicted probabilities of rearrest, but the
qualitative message about benefits for certain persons with
these characteristics does not change.
It is important to note that we do not estimate separate
models for each combination of characteristics. Instead,
the multivariate analysis uses all of these characteristics
in one model simultaneously, while allowing prediction of
the relevant probabilities under various hypothetical sce-
narios, that is, using various combinations of these charac-
teristics. More detailed discussion of the regressions and
simulations will be available from the authors in a forth-
coming technical report.
10

while the rearrest rate for discretionary parolees was ing that discretionary parolees were selected for release
four percentage points lower (57 percent; table 3).25 based not only on their criminal histories (for which our
model controlled) but also on individual attitude, moti-
It is notable that mandatory parolees, who account for vation, and preparedness (which the parole boards took
the largest share of released prisoners, fare no better with into consideration but could not be controlled for in our
supervision than similar prisoners released without model). One would expect that discretionary parolees are
supervision. While discretionary parolees are less likely to better positioned to succeed than the rest of the released
be rearrested, this difference is relatively small consider- prisoner population because they have met the parole
board’s selection criteria. Accordingly, supervision may
not be the chief reason for this difference in outcomes.
TABLE 3. Predicted Probability of Rearrest Two Years after
Release, by Supervision Status at 1994 Release
FOR WHOM DOES SUPERVISION
Unconditional Mandatory Discretionary
MATTER MOST?
releasees (%) parolees (%) parolees (%)
While the modeling revealed small overall differences in
OVERALL 61 61 57 rearrest outcomes based on supervision status, certain
Male 60 62 58 subgroups were predicted to have substantial reductions
Female 67 51 51
in recidivism when supervised. For example, holding all
Black 68 67 61
other characteristics constant, the predicted probability
Non-black 54 56 53
of rearrest for a discretionary parolee with few prior
Low release age 61 60 57
arrests was nine percentage points lower than for an
Medium release age 62 62 58
unconditional releasee with a similar criminal history,
High release age 52 53 48
and five percentage points lower than for a mandatory
Few prior arrests 53 49 44
parolee (table 3). The likelihood of rearrest for a dis-
Medium prior arrests 59 57 52
cretionary parolee who had been serving time for a
High prior arrests 68 70 66
technical violation was eight percentage points lower
Violent offense 55 56 55
Property offense 68 67 62
than for a similar unconditional releasee, and five per-
Drug offense 56 61 54 centage points lower than for a mandatory parolee.
Public order and other
offense 65 57 55 The impact of supervision seems to differ based on the
New sentence 56 58 54 offense type. Notably, only parolees incarcerated for pub-
Revocation + new lic order or other offenses were predicted to have lower
sentence 59 62 53 rearrest rates—eight to ten percentage points lower—
Revocation (technical) 71 68 63 than their unconditional release counterparts. We dis-
Low resource cuss findings by offense type in more detail below, in the
deprivation 59 61 56 section “How Does Supervision Affect the Largest
Medium resource Release Groups?”
deprivation 61 61 57
High resource The most sizable difference in outcomes was for females:
deprivation 62 62 58 the predicted probability of rearrest for a female parolee
Source: Urban Institute analysis of Bureau of Justice Statistics data. See (discretionary and mandatory) was 16 percentage points
“Methodology” sidebar for definitions of characteristics. lower than for a female released unconditionally.

We took this analysis a step further by combining all of
the factors included in the modeling to predict the prob-
abilities of rearrest for mandatory and discretionary
parolees and unconditional releasees with each combi-
nation of characteristics. For example, we estimated re-
arrest outcomes by supervision status for a young, black
male with few prior arrests who is a new court commit-
ment for a drug offense and returned to an area with
high resource deprivation. By comparing the differences
in predicted rearrest rates for each combination of char-
acteristics, we were able to gauge the effect of supervision
status on different groups of releasees—that is, to statis-
tically address the question “For whom does supervision
matter most?”
Some combinations of characteristics yielded very high
benefits from supervision. In other words, their expected
rearrest rates were significantly lower when released to
supervision than unconditionally released without it.
Other groups, by contrast, yielded small or even negative
effects from supervision. However, very few people
released in 1994 had the combinations of characteristics
that were expected to achieve the highest—or lowest—
benefits from supervision. In fact, the typical releasee
yielded small, if any, gains from supervision. In this sec-
tion, we describe the characteristics of the highest and
lowest benefiting groups, then turn to a description of
predicted rearrest rates by supervision status for the
largest shares of released prisoners.
Who Might Benefit from Supervision the Most?
The predicted probability of rearrest for some male dis-
cretionary parolees was as much as 20 percentage points
lower than that of male unconditional releasees with the
same characteristics. These “high-benefiting” males
tended to be black, had few prior arrests, were serving
time for parole or probation revocations, and were con-
victed of “other” offenses—mostly public order offenses.
Conducting the same analyses for females showed simi-
lar patterns, with some female discretionary parolees
having predicted rearrest rates as much as 34 percentage
11
points lower than their unconditional counterparts. No
patterns were evident among high-benefiting males or
females with regard to age or resource deprivation in the
communities to which they returned. Repeating the
analysis for males and females to compare mandatory
parolees with unconditional releasees showed a similar
set of characteristics among the highest benefiting
groups, although the race trend was not as strong and the
potential reduction in rearrest rates was not as high. In
sum, those who appear to benefit most from supervision
are low-risk, low-level offenders, who account for small
shares of the overall release cohort. These individuals are
possibly more responsive to the sanctions and services
provided by supervision, given their minimal prior
involvements with the justice system.
Who Might Benefit from Supervision the Least?
Some male discretionary parolees achieved little to no
benefit from supervision in terms of recidivism out-
comes. These males tended to be white, had high num-
bers of prior arrests, were serving time for new court
commitments, and were convicted of violent or drug
offenses. In fact, some males with these characteristics
had a predicted probability of rearrest roughly equal to
or as much as 10 percentage points higher than uncondi-
tional releasees with the same characteristics. The female
discretionary parolees who benefited least from super-
vision were similar to their male counterparts, although
no female discretionary parolees had higher rearrest rates
than their unconditional counterparts. As with the high-
est benefiting groups, no patterns emerged for age or
resource deprivation in the communities to which the
lowest benefiting individuals returned. Comparing
mandatory parolees with unconditional releasees
revealed those lowest benefiting individuals to also be
white, have high numbers of prior arrests, have been
serving time for new court commitments, and have been
convicted of violent or drug offenses, although the
potential negative impact of supervision on males was
greater than for females. These results indicate that some
of the higher rate, more serious offenders may in fact
12
benefit the least from supervision. It may be that these
individuals, who have extensive and serious criminal his-
tories, are immune to the deterrent effect of supervision
and unthreatened by the possibility of reincarceration.
How Does Supervision Affect the
Largest Release Groups?
The highest and lowest benefiting groups described
above account for very small shares of the total release
population. Conversely, male drug, property, and violent
offenders together account for over 80 percent of the
release cohort in 1994. We therefore sought to address
whether those individuals who account for sizable shares
of the population released from prison are predicted to
have lower rearrest rates when supervised after release. In
short, supervision impacts rearrest outcomes differently
based on the incarcerating offense type.
Specifically, supervision does not play much of a role
among those incarcerated for a violent offense (roughly
one-fifth of the released population). Even within this
group, when assessing the effects that supervision has on
subgroups based on age and gender, we find little evi-
dence of differences in the predicted probability of rear-
rest based on the three release mechanisms analyzed
here. Discretionary parole does seem to benefit property
offenders (roughly one-third of the released population),
although predicted rearrest rates for mandatory parolees
are virtually the same as for unconditional releasees.
On the other hand, the predicted rearrest rates for drug
offenders are the same for discretionary parolees and
unconditional releasees, while mandatory parolees actu-
ally have higher rearrest rates than the other two groups
(table 4). Mandatory parolees may have higher rearrest
rates because, unlike their discretionary counterparts,
they are a higher-risk population; unlike the uncondi-
tional releasees, they are subject to heightened surveil-
lance, which may include frequent drug testing.
In sum, for the largest release groups, supervision is asso-
ciated with lower rearrest rates only among property
offenders released via discretionary parole. Among drug
offenders, mandatory release actually predicts higher
rearrest outcomes.
WHEN DOES SUPERVISION
MATTER MOST?
According to the BJS recidivism study, among all prison-
ers who recidivated within three years of release, nearly
two-thirds recidivated in the first year. In our sample,
unconditional releasees were, on average, rearrested the
earliest (9.9 months)—about half a month before
mandatory parolees (10.4 months) and a month and
a half before discretionary parolees (11.5 months).
However, the likelihood of rearrest for each release
group changed over time. Controlling for all other
characteristics, discretionary parolees had a 24 percent
likelihood of being rearrested in the first six months after
release, compared with 27 percent of both mandatory
parolees and unconditional releasees. Discretionary
parolees who had not been rearrested in the first six
months after release (“survivors”) were less likely to be
rearrested in the next six months to one year after release
than mandatory parolees and unconditional releasees.
Between 18 and 24 months after release, the likelihood of
rearrest for survivors was roughly equivalent among all
three groups (figure 4).
It is possible that these findings reflect the fact that per-
sons released to supervision were not necessarily super-
vised throughout the follow-up period. Since data on
the length of supervision for our sample were not avail-
able, we chose to measure outcomes at two years after
release—the average length of supervision nationally at
that time. Over time, however, fewer and fewer people
may have been on supervision, and by the end of the
two years, individuals in the study may have had similar
supervision statuses and therefore similar recidivism
rates. In sum, while discretionary parolees were less likely
to be rearrested initially, the benefits of (discretionary)
supervised release dropped systematically over time.
 13

TABLE 4. Predicted Probability of Rearrest Two Years after Release for Largest Release Groups, by Supervision Status at 1994 Release

Predicted probability of rearrest (%)

Share of 1994
release cohort (%) Unconditional releasees Mandatory parolees Discretionary parolees

Property offenders
Young males 11.7 68 67 62
Medium males 8.5 68 68 63
Older males 10.5 59 60 53

Drug offenders
Young males 10.7 55 61 54
Medium males 8.0 55 62 55
Older males 9.7 45 54 45

Violent offenders
Young males 8.5 55 56 56
Medium males 5.4 55 58 57
Older males 7.4 45 49 47

Source: Urban Institute analysis of Bureau of Justice Statistics data. See “Methodology” sidebar for definitions of age groupings.

CONCLUDING OBSERVATIONS
FIGURE 4. Probability of Rearrest, by Supervision Status
at Release We originally hypothesized that prisoners released to
Percent supervision would recidivate at lower rates than prison-
30 ers released without supervision. We expected lower
rearrest rates because of the characteristics of supervised
25 releases (on average, lower risk than unsupervised
releases) and the presumed deterrent effect of supervi-
Mandatory
parolees
sion. Our findings did confirm that certain parolees—
20 Discretionary
parolees those released by a parole board or other authority—had
criminal histories indicating a lower risk. Mandatory
15 parolees, on the other hand, are more similar to uncon-
ditional releasees than to discretionary parolees. Upon
Unconditional
10 reflection, this similarity is not surprising given that
releases
mandatory parolees and unconditional releasees are
both released from prison on a predetermined date,
5
without any form of screening to determine readiness
for release.
0
0–6 7–12 13–18 19–24
months months months months Our recidivism findings also differed from our original
Source: Urban Institute analysis of Bureau of Justice Statistics data. hypothesis. On the whole, discretionary parolees were
14

STATE VARIATION
TABLE 5. Share of Prisoners Released Conditionally and
While this research brief focuses on national-level trends, Unconditionally in 1994, by State
it is important to note that the use, duration, and intensity
Unconditional Mandatory Discretionary
of postrelease supervision varies significantly across State releasees (%) parolees (%) parolees (%)
states.1 While postprison supervision is implemented
Arizona 16 1 83
differently across states, it generally involves a set of
California 2 97 1
conditions such as abstinence from drugs, maintaining
Florida 22 1 77
employment, observing curfews, and staying away from cer-
Illinois 2 98 0
tain high-risk places and persons. Enforcement of those Maryland 9 49 42
conditions may include home visits, drug testing, electronic Michigan 9 0 91
monitoring, and even Global Positioning System satellites, Minnesota 2 77 21
where individuals’ movements are tracked 24 hours a day. New Jersey 22 0 78
New York 5 13 82
The share of prisoners released to parole supervision North Carolina 3 59 38
varies considerably by state as well. In Ohio, more than Ohio 39 0 61
one-third of the prisoners released in 1994 were not sub- Oregon 0 35 65

ject to any postprison supervision. By contrast, in Illinois Texas 2 40 58

Virginia 3 53 44
nearly all prisoners were released to supervision following
mandatory release (table 5). In most states, the conditions Source: Urban Institute analysis of Bureau of Justice Statistics data.
These statistics exclude the category of releases called
of supervision are similar for both discretionary and
“transfer/other.”
mandatory parolees, although the length of time on super-
vision often varies (on average, discretionary releases
spend more time on supervision than mandatory
sons released to supervision have more extensive criminal
releases2). In Maryland, for example, the average time
histories than their unsupervised counterparts and may be
on supervision for a prisoner released by a parole board
more likely to recidivate as a result. Other states have dif-
in 1994 was almost three years (35 months). By contrast,
ferent policies, which result in the highest risk prisoners as
prisoners released to supervision by mandatory release
the most likely to serve their full sentence behind bars.
were supervised in the community for less than a year on
Note that many states have undergone substantial changes
average (9 months).3 In other states, the supervision
in their release policies and supervision practices since
period for both types of parolees is similar.
1994, which may alter the current distribution of releasees
by supervision type and affect recidivism rates.
Recidivism outcomes for the various types of releasees
also vary by state. In about two-thirds of the states included
1
in the study, discretionary parolees are less likely to be Anne Piehl and Stefan LoBuglio. Forthcoming. “Does Supervision
Matter?” In Prisoner Reentry and Crime in America, edited by Jeremy
rearrested than either unconditional releasees or mandatory
Travis and Christy Visher. Cambridge, U.K.: Cambridge University Press.
parolees. In other states, the outcomes are reversed, with
2
Timothy Hughes, Doris Wilson, and Allen Beck. 2001. “Trends in State
either mandatory parolees or unconditional releasees being
Parole, 1990–2000.” Bureau of Justice Statistics Special Report. Wash-
least likely to be rearrested. These differences in outcomes ington, DC: U.S. Department of Justice, Office of Justice Programs.
are likely due, at least in part, to variations in state policies 3
Maryland Department of Public Safety and Correctional Services,
on who is supervised after release. In some states, per- Office of Planning and Statistics.

less likely to be rearrested than unconditional releasees in
the two years after release, but the rearrest rates for
mandatory parolees and unconditional releasees were
very similar. When we expanded our analysis to isolate
the effect of supervision on rearrest, independent of
demographic characteristics and known risk factors such
as criminal histories, the difference in rearrest rates
decreased even further: the predicted probability of re-
arrest for mandatory parolees and unconditional
releasees was identical at 61 percent, while the probabil-
ity of rearrest for discretionary parolees was only 4 per-
centage points lower at 57 percent. In other words,
mandatory parolees, who represent the largest share of
released prisoners, are no better off in terms of rearrests
than prisoners released without supervision. Those
screened by a parole board are less likely to be rearrested,
but the difference was relatively small, particularly given
that parole boards selected these individuals as low-risk
candidates for release. Moreover, even this modest dif-
ference may be due to factors other than supervision,
given that parole boards base their decisions on such
factors as attitude, motivation, and preparedness for
release that our model cannot take into account.
The modest difference in overall rearrest rates should not
suggest that supervision had no effect on rearrests for any-
one in any way at any time. For certain individuals—
including females, those with few prior arrests, public
order offenders, and technical violators—being released to
supervision, especially via discretionary release, predicted
rearrest rates as much as 16 percentage points lower than
rates observed with unconditional release. In addition,
persons with combinations of these characteristics (e.g.,
females with few prior arrests who were incarcerated for
a public order offense) who were released to supervision
were predicted to have even lower rearrest rates. On the
other hand, certain high-rate offenders—such as white
males with many prior arrests who were serving time for
new court commitments for violent or drug offenses—
received no benefit in terms of reduced rearrest rates from
supervision.26 In fact, some of these males had a higher
15
predicted probability of rearrest than similar individuals
released without supervision. Notably, few prisoners have
a combination of characteristics that yield either the high-
est or lowest benefits from supervision.
By contrast, the public safety impact of supervision is
minimal and often nonexistent among the largest shares
of the release cohort—male property, drug, and violent
offenders. Supervision does not appear to improve
recidivism outcomes for violent offenders or property
offenders released to mandatory parole. Rather, our
analysis shows that supervision is only associated with
lower rearrest rates among discretionary parolees who
had been incarcerated for a property offense. In fact, for
male drug offenders, mandatory release to supervision
predicts higher rearrest rates than for unconditional
releasees or discretionary parolees. These higher rates
may reflect the fact that mandatory parolees are a
higher-risk population than discretionary parolees
and face heightened surveillance (such as drug testing)
compared with unconditional releasees. In short,
while postprison supervision may have modest effects
on recidivism in some cases, it does not appear to
improve rearrest rates for the largest subsets of re-
leased prisoners.
It bears repeating that the nature of our analysis does not
allow for insights into whether certain types of supervi-
sion, such as neighborhood-based or case management
models, are more effective than others or whether there
are differences in outcomes across states. It is also unclear
how much rearrest outcomes are the result of policy
directives (e.g., a decision to watch more closely and
arrest more quickly) and not criminal activity alone. At
the same time, given our country’s heavy reliance on
parole to manage those released from prison, it is dis-
couraging—although not wholly unexpected—to find
that the overall effect of supervision appears to be mini-
mal. For years, parole experts have suspected that parole
supervision was ineffective although national data did not
exist to support those assumptions. At a 1998 meeting on
16
supervision sponsored by the Department of Justice,
experts described current probation and parole models in
a state of “dangerous opportunity,” lacking clarity in pur-
pose as well as public and political credibility.27
More recently, reentry experts reached similar, if bolder,
conclusions. Jeremy Travis, preeminent scholar of pris-
oner reentry and current president of John Jay College
of Criminal Justice, has called for “an end to parole
as we know it.”28 Travis has conceptualized a new
approach to supervision, recommending innovative
incentives for early release and limited supervision con-
ditions that would align with each prisoner’s needs and
risks.29 Martin F. Horn, commissioner of the New York
City Departments of Corrections and Probation and
former corrections secretary for Pennsylvania, has pro-
posed abolishing parole altogether given the lack of evi-
dence that it discourages criminal behavior. Instead of
parole, he recommends that released prisoners be pro-
vided with vouchers that can be used for transitional
services that they choose.30 Joan Petersilia, a professor
at the University of California, Irvine, and renowned
researcher on parole, has argued for the reinstitution31
and redesign of discretionary parole, relying more
heavily on risk assessment tools that predict a prisoner’s
likelihood of committing future crimes in making
release decisions.32
Prior research and discussion have suggested several
reasons why parole, as typically implemented, is not as
effective as it could be.33

Parole supervision is, in fact, quite minimal in most
cases. Most parole officers manage large caseloads
(an average of 70 parolees apiece) and typically meet
with individuals for about 15 minutes once or twice a
month.34 Why would we expect such a small amount
of contact to make a large amount of difference?35
Parolees don’t: According to one study of parolees,
most report that their parole officer did not have a
major positive or negative impact on their postprison
behavior.36 Clearly parole supervision must be more
than occasional if it is to have an appreciable effect.

Parole officers are often located far from the neigh-
borhoods where parolees reside, and therefore lack
an understanding of the situational context that geo-
graphically oriented supervision could provide.
Similar to community policing, community-based
parole officers could get to know their neighborhood
resources and high-risk areas, and thus be in a better
position to meaningfully assist and sanction parolees
on their caseloads.37

In most states, responses to violations are often in-
consistent and inappropriate to the seriousness of the
infraction.38 Parolees may violate conditions without
being caught or may be caught several times but
receive nothing more than a warning, and then a
seemingly random violation results in their return
to prison for the remainder of their sentence.39 The
research literature suggests that to be effective, punish-
ment should be immediate and predictable, with clear,
enforceable consequences for violations.40 Parole could
benefit from an array of intermediate sanctions to
employ in response to violations, as opposed to the
“all or nothing” approach often used today. Such an
overhaul in the parole violation and revocation
process could, ideally, enhance the deterrent effect of
supervision.

In recent years, the parole function has shifted from a
service orientation to a surveillance-oriented, control-
based strategy centered on monitoring behavior,
detecting violations, and enforcing the rules.41 New
surveillance technologies such as drug testing, elec-
tronic monitoring, and Global Positioning System
satellites make it easier and more efficient to monitor
behavior than traditional casework.42 However, prior
studies indicate that surveillance alone will not invoke
change. Rather, a mix of appropriate43 treatment and
surveillance is needed to positively affect offender

behavior.44 Importantly, the field’s major evaluation of
Intensive Supervision Programs found that supervi-
sion can effectively direct individuals to treatment and
community programming.45 This positive finding
should inform new efforts to improve parole.

It is also possible that the traditional approach to
parole supervision is conceptually ill-suited to reduce
recidivism among released prisoners. As currently
implemented, supervision either focuses on ex-
prisoners’ risks, through a control model, or on their
needs, through a support model. Researchers Shadd
Maruna and Tom LeBel suggest that a strengths-based
approach that builds on an ex-prisoner’s positive
assets would be perceived as more legitimate by ex-
offenders and would be more effective in allowing
them to take responsibility and become part of the
community.46
Looking Forward
Given this country’s large-scale investment in supervi-
sion—it is currently the most prevalent tool in managing
reentry—the topic warrants additional research atten-
tion and should be brought to the forefront of every pol-
icy discussion on the topic of prisoner reentry. It is
critical to understand not only why supervision does not
work as well as it should across the board, but also why
supervision does work for some groups and how similar
gains could be realized for larger subsets of the parole
caseload. Importantly, correctional leaders need analysis
that gets inside “the black box” and considers the differ-
ent models of parole and their relationship to recidivism
outcomes. In addition, it is worth considering whether
any lessons from the discretionary release process
could be transferred to postrelease supervision.
That is, are there other ways to stimulate good behavior,
increase individual motivation, and better prepare a
greater share of prisoners prior to release? Moreover,
could incentives be put in place to enable individuals to
earn their way off of parole supervision? There should
also be more testing and evaluating of innovative super-
17
vision strategies across the country. Given the diversity
of practice across the states as well as the experimenta-
tion around prisoner reentry currently under way, there
is a ripe opportunity to assess what is going on and learn
from the field.
It is important to note that despite disappointing find-
ings and substantial criticism from corrections experts
and the public alike, few would recommend postprison
supervision be abandoned altogether. As is evident from
our study, prisoners released unconditionally are also
highly likely to reoffend upon release. Further, common
sense suggests that prisoners, especially the high-risk pris-
oners that supervision appears least likely to help, war-
rant some sort of structure—a mix of supervision and
support—after prison. As a nation, we face an oppor-
tunity to rethink, revise, and perhaps reinvent parole
supervision so that it is vastly better at producing public
safety outcomes and enhancing the odds of successful
reintegration for the more than 600,000 individuals
leaving prison each year. Without renewed efforts to
improve the public safety benefits of postprison super-
vision, our reliance on parole serves little purpose apart
from providing false comfort.
ENDNOTES
1 Bureau of Justice Statistics. Reentry Trends in the U.S. Available at:
http://www.ojp.usdoj.gov/bjs/reentry/releases.htm.
2 See Joan Petersilia. 2003. When Prisoners Come Home: Parole and
Prisoner Reentry. New York: Oxford University Press; Edward Rhine.
1997. “Probation and Parole Supervision: In Need of a New Narrative.”
Corrections Management Quarterly 1(2): 71–75.
3
Lauren Glaze and Sera Palla. 2004. “Probation and Parole in the
United States, 2003.” Bureau of Justice Statistics Bulletin. Washington,
DC: U.S. Department of Justice, Office of Justice Programs.
4 Timothy Hughes, Doris Wilson, and Allen Beck. 2001. “Trends in
State Parole, 1990–2000.” Bureau of Justice Statistics Special Report.
Washington, DC: U.S. Department of Justice. http://www.ojp.usdoj.
gov/bjs/pub/pdf/tsp00.pdf.
18
5 unconditionally, estimated at 13 percent nationwide in 1994, is under-
Ibid. The study also found that more than half of discretionary
parolees successfully complete their term of supervision compared
with one-third of mandatory parolees.
6
Patrick Langan and David Levin. 2002. “Recidivism of Prisoners
Released in 1994.” Bureau of Justice Statistics Special Report.
Washington, DC: U.S. Department of Justice, Office of Justice
Programs.
7
Ibid.
8 In 2000, almost 20 percent of state prisoners—or 112,000 persons—
were released from state prisons without any postprison supervision.
See Hughes, Wilson, and Beck, “Trends in State Parole, 1990–2000.”
9
Ibid. Critics of discretionary release claim that abolishing parole
would both reduce disparities in length of time served and keep
inmates imprisoned for longer amounts of time. In fact, research has
shown that length of stay is actually longer in states that still utilize
discretionary parole.
10
Ibid. The percentages shown in figure 1 represent the shares of all
prisoners released each year who were released by these three methods.
The total pool of released prisoners also includes releases to probation,
commutations, and other unspecified releasees.
11 Ibid.
12
For more information see table 3.1 in Petersilia, When Prisoners
Come Home: Parole and Prisoner Reentry
13 Hughes, Wilson, and Beck, “Trends in State Parole, 1990–2000.”
14 In fact, while uncommon, in states that have retained discretionary
parole, mandatory releasees may have been denied early release by a
parole board.
15 Hughes, Wilson, and Beck, “Trends in State Parole, 1990–2000.”
16 While unconditional releasees accounted for around 40 percent of
prisoners released during the first half of the 20th century, the percent-
age of prisoners released without supervision declined steadily until the
1980s. Between 1980 and the mid-1990s, the share of prisoners released
unconditionally fluctuated between 11 and 17 percent, and has risen
to just under 20 percent since then. Because of the growth in prison
releases over the century, the number of prisoners released uncondi-
tionally has continued to rise over time. Jeremy Travis and Sarah
Lawrence. 2002. Beyond the Prison Gates: The State of Parole in America.
Washington, DC: The Urban Institute. http://www.urban.org/
UploadedPDF/310583_Beyond_prison_gates.pdf.
17
It is worth noting that while the 15-state BJS sample represents two-
thirds of all prisoners released in 1994, the portion of prisoners released
represented. Jodi M. Brown, Darrell K. Gilliard, Tracy L. Snell, James J.
Stephan, and Doris James Wilson. 1996. Correctional Populations in the
United States, 1994. Washington, DC: U.S. Department of Justice,
Bureau of Justice Statistics.
18 Data on the amount of time individuals in our sample actually
served under parole supervision were not available. We therefore
chose to measure outcomes at two years after release—the approxi-
mate time served on parole nationally in the 1990s (23 months in 1990;
26 months in 1999 per Hughes, Wilson, and Beck, “Trends in State
Parole, 1990–2000”).
19
For example, other studies have found that younger offenders, those
convicted of property offenses, and those with more extensive criminal
histories are more likely to reoffend. Langan and Levin, “Recidivism
of Prisoners Released in 1994”; Christy Visher, Vera Kachnowski,
Nancy G. La Vigne, and Jeremy Travis. 2004. “Baltimore Prisoners’
Experiences Returning Home.” Washington, DC: The Urban Institute.
http://www.urban.org/url.cfm?ID=310946.
20It is unsurprising that unconditional releasees serve longer terms in
prison given that—by definition—they are not released early like their
discretionary and mandatory counterparts, and therefore serve their
entire sentence in prison.
21Langan and Levin, “Recidivism of Prisoners Released in 1994.”
Note that we report recidivism outcomes for a two-year follow-up
period because it most closely represents the average length of time ex-
prisoners are under supervision in the community, and is thus the best
point of comparison in assessing the impact of supervision on re-
cidivism outcomes. In 1990, the average time served on parole was
23 months; in 1999, it was 26 months (Hughes, Wilson, and Beck,
“Trends in State Parole, 1990–2000”).
22
For more discussion, see Anne Piehl and Stefan LoBuglio. Forth-
coming. “Does Supervision Matter?” In Prisoner Reentry and Crime in
America, edited by Jeremy Travis and Christy Visher. Cambridge, U.K.:
Cambridge University Press. Also see “Data Sources and Limitations”
sidebar.
23 In the two years after release, 38 percent of unconditional releasees
were reconvicted, as were 37 percent of mandatory parolees and 32 per-
cent of discretionary parolees. Twenty percent of mandatory parolees
returned to prison for a new offense, as did 17 percent of unconditional
releasees and 15 percent of discretionary parolees.
24
As noted earlier, discretionary and mandatory parole supervision do
not differ from each other systematically across states.
25Because California often influences the national trends so heavily, we
re-ran our model including all study states except California (and

Delaware, which was excluded from the entire analysis, as discussed in
the “Data Sources and Limitations” sidebar). The overall recidivism
findings change when California is excluded, but not dramatically: the
predicted probability of rearrest for unconditional releasees rises to
63 percent, compared with 60 percent for mandatory parolees and
56 percent for discretionary parolees.
26This finding may seem at odds with the treatment literature, which
suggests that treatment is most effective when targeting the crimino-
genic needs of high-risk offenders. While our research suggests that
supervision benefits low-risk ex-prisoners more than their high-risk
counterparts (in terms of rearrest outcomes), our study does not speak
to the effect of treatment (or the mix of treatment and supervision) on
either high- or low-risk populations.
27
The focus group participants determined five possible outcomes for
the future of parole and probation, ranging from continuing the cur-
rent state of affairs by “muddling along” to a model explicitly focused
on public safety. Walter Dickey and Michael Smith. 1998. Dangerous
Opportunity: Five Futures for Community Corrections. Washington, DC:
U.S. Department of Justice, Office of Justice Programs. http://www.
ojp.usdoj.gov/probation/.
28
Jeremy Travis. 2002. “Thoughts on the Future of Parole.” Remarks
delivered at the Vera Institute of Justice, New York, May 22. http://
www.urban.org/UploadedPDF/410521.pdf.
29
Jeremy Travis. Forthcoming. But They All Come Back: Facing the
Challenges of Prisoner Reentry. Washington, DC: Urban Institute Press.
30
Martin F. Horn. 2001. “Rethinking Sentencing.” Corrections Man-
agement Quarterly 5(3): 34–40.
31
Reinstitution of discretionary parole would occur in the 16 states that
have abolished it.
32
Petersilia, When Prisoners Come Home: Parole and Prisoner Reentry.
33
Some of the major issues and recommendations summarized here are
discussed in more detail in the Report of the Reentry Policy Council.
The report consists of policy statements and recommendations that
address the various dimensions of prisoner reentry. It also includes
examples of innovative parole strategies currently being implemented
around the country. The report reflects the broad consensus of the Re-
Entry Policy Council, a bi-partisan group of leading policymakers and
practitioners representing a broad spectrum of criminal justice, health,
housing, and employment systems. The Report of the Re-Entry Policy
Council is available at http://www.reentrypolicy.org.
34 Ibid.
19
35
At the same time, research suggests that more intensive supervision
results in more violations, but not necessarily more public safety bene-
fits. Joan Petersilia and Susan Turner. 1993. “Evaluating Intensive
Supervision Probation/Parole: Results from a Nationwide Experi-
ment.” NIJ Research in Brief. Washington, DC: U.S. Department of
Justice, Office of Justice Programs.
36
Edward Zamble and Vernon Quinsey. 1997. The Criminal Recidivism
Process. Cambridge: Cambridge University Press.
37
Petersilia, When Prisoners Come Home: Parole and Prisoner Reentry.
38
National Institute of Corrections. 2004. Parole Violations Revisited:
A Handbook on Strengthening Parole Practices for Public Safety and
Successful Transition to the Community. Washington, DC: U.S.
Department of Justice, National Institute of Corrections. NIC 019833.
http://www.nicic.org/pubs/2004/019833.pdf.
39
Ibid.
40
Peggy Burke. 1997. Policy-Driven Responses to Probation and Parole
Violations. Silver Spring, MD: Center for Effective Public Policy.
41 Petersilia, When Prisoners Come Home: Parole and Prisoner Reentry.
42 The intensive supervision literature indicates that when parolees are
monitored more closely, they are caught more frequently, although the
public safety benefits are unclear. Petersilia and Turner, “Evaluating
Intensive Supervision Probation/Parole: Results from a Nationwide
Experiment.”
43
In-prison treatment interventions are most effective when programs
are matched to prisoners’ risks and needs, when they are well-managed,
and when the intervention is supported through postrelease super-
vision. Gerald Gaes, Timothy Flanagan, Laurence Motiuk, and Lynn
Stewart. 1999. “Adult Correctional Treatment.” In Prisons, edited by
Michael Tonry and Joan Petersilia. Chicago: University of Chicago
Press.
44
Lawrence W. Sherman, Denise Gottfredson, Doris MacKenzie, John
Eck, Peter Reuter, and Shawn Bushway. 1998. “Preventing Crime:
What Works, What Doesn’t, What’s Promising.” Research in Brief.
Washington, DC: National Institute of Justice. http://www.ncjrs.org/
works/.
45 Joan Petersilia. 1998. “A Decade of Experimenting with Intermediate
Sanctions: What Have We Learned?” Federal Probation 62(2): 3–9.
46
Shadd Maruna and Thomas P. LeBel. 2003. “Welcome Home?
Examining the ‘Reentry Court’ Concept from a Strength-based
Perspective.” Western Criminology Review 4(2): 91–107.
20
ACKNOWLEDGMENTS
The authors would like to thank several individuals
who contributed to this report. We are grateful to
Anne Piehl of the Kennedy School of Government;
Bill Sabol of the U.S. Government Accountability
Office; Joan Petersilia of the University of
California, Irvine; and Edward Rhine of the Ohio
Department of Rehabilitation and Correction for
methodological advice and valuable comments on
drafts of this report. From the Urban Institute’s
FOR FURTHER READING
Lattimore, Pamela, Susan Brumbaugh, Christy Visher,
Christine Lindquist, Laura Winterfield, Meghan Salas,
and Janine Zweig. 2004. “National Portrait of SVORI.”
Washington, DC: The Urban Institute. http://www.
urban.org/url.cfm?ID=1000692.
Lawrence, Sarah, and Jeremy Travis. 2004. “The New
Landscape of Imprisonment: Mapping America’s Prison
Expansion.” Research Report. Washington, DC: The
Urban Institute. http://www.urban.org/url.cfm?
ID=410994.
Lawrence, Sarah, Daniel Mears, Glenn Dubin, and
Jeremy Travis. 2002. “The Practice and Promise of
Prison Programming.” Research Report. Washington,
DC: The Urban Institute. http://www.urban.org/url.
cfm?ID=410493.
Lynch, James, and William Sabol. 2001. “Prisoner
Reentry in Perspective.” Crime Policy Report, vol. 3.
Washington, DC: The Urban Institute. http://www.
urban.org/url.cfm?ID=410213.
Solomon, Amy L., Kelly Dedel Johnson, Jeremy Travis,
and Elizabeth McBride. 2004. “From Prison to Work.”
Washington, DC: The Urban Institute. http://www.
urban.org/url.cfm?ID=411097.
Solomon, Amy L., Michelle Waul, Asheley Van Ness, and
Jeremy Travis. 2004. “Outside the Walls: A National
Snapshot of Community Reentry Programs.” Wash-
ington, DC: The Urban Institute. http://www.urban.
org/url.cfm?ID=410911.
Justice Policy Center, Christy Visher provided sage
advice throughout the life of the study; Nancy
La Vigne and Jeremy Travis gave us critical feed-
back at early stages of the project; and Elizabeth
McBride contributed her editorial talent on the
final drafts. This report was made possible
through the generous support of the JEHT
Foundation. We extend a special thanks to Scott
Bane for his substantive, important questions—
and his patience in waiting for the answers.
Travis, Jeremy, and Sarah Lawrence. 2002. “Beyond the
Prison Gates: The State of Parole in America.” Wash-
ington, DC: The Urban Institute. http://www.urban.
org/url.cfm?ID=310583.
Travis, Jeremy, Elizabeth M. Cincotta, and Amy L.
Solomon. 2003. “Families Left Behind: The Hidden
Costs of Incarceration and Reentry.” Washington, DC:
The Urban Institute. http://www.urban.org/url.cfm?
ID=310882.
Travis, Jeremy, Amy L. Solomon, and Michelle Waul.
2001. “From Prison to Home: The Dimensions and
Consequences of Prisoner Reentry.” Washington, DC:
The Urban Institute. http://www.urban.org/pdfs/from_
prison_to_home.pdf
Visher, Christy, Nancy La Vigne, and Jill Farrell. 2003.
“Illinois Prisoners’ Reflections on Returning Home.”
Washington, DC: The Urban Institute. http://www.
urban.org/url.cfm?ID=310846.
Visher, Christy, Vera Kachnowski, Nancy La Vigne, and
Jeremy Travis. 2004. “Baltimore Prisoners’ Experiences
Returning Home.” Washington, DC: The Urban
Institute. http://www.urban.org/url.cfm?ID=310946.
To receive free monthly email updates on the research of the
Justice Policy Center, join the Center’s email distribution
list by sending an email to JPC@ui.urban.org.
The views expressed are those of the authors and should
not be attributed to the Urban Institute, its trustees, or its
funders.